The Detection and Exploration of Planets from the Trans-atlantic Exoplanet Survey

The Detection and Exploration of Planets from the Trans-atlantic Exoplanet Survey Thesis by Francis T. O Donovan In Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy California Institute of Technology Pasadena, California 2008 (Defended July 31, 2007)

ii c 2008 Francis T. O Donovan All Rights Reserved

iii For Bridget and Fran, now united with Francis, Jackie, Bríd, James, and Nellie, For my family of PhDs: Tom, Vera, and Bridget, For Cathy, a ridiculously lovely human being, And for all who believed that I could.

iv Acknowledgements As those most likely to actually read my entire thesis, I would first like to thank the faculty who served on my candidacy and defense committees. Thanks to Wal Sargent and Andrew Blain for stepping in on short notice, to Tony Readhead for understanding the lure of extrasolar planets, to Mike Brown for putting Pluto in its place, and to Re em Sari for helping to keep Cathy alive under water! This thesis work could not have succeeded without Lynne Hillenbrand, who provided all of the benefits of having two advisors without making me try to please two masters. Finally, I owe a lot to my advisor David Charbonneau, who first enticed me to the dark side of exoplanets, and kept my spirits up when the prospects of planet discovery looked bleak. Though he moved 3,000 miles away to be with his wife the nerve! he proved that long-distance advising can work. (Indeed, I still cringe at the sound of our office phone.) Strange though it is to acknowledge an inanimate object, without the tireless workhorse that is Sleuth, I would not have helped discover three planets, which might have put the brakes on graduating. Despite a lonely existence on Mount Palomar, Sleuth was very reliable, even when rained on, unless David or I got on a plane! More importantly, I must thank those at Palomar Observatory, especially Dipali, Jean, Karl, and Rose, who kept me company on my many one-hour visits. I sincerely thank Robert Brucato, Michael Doyle, Karl Dunscombe, Richard Ellis, Brian Gordon, John Henning, Linley Kroll, Steven Kunsman, Jean Mueller, Hal Petrie, Andrew Pickles, Nick Scoville, Merle Sweet, Robert Thicksten, Greg van Idsinga, Richard Wetzel, and Daniel Zieber for their assistance with the fabrication, operation, and maintenance of the Sleuth instrument. My discoveries with Sleuth would never have happened without the help of the

v TrES team and collaborators. I have never learned so much about astronomy as I did at our May meetings. My thanks to Roi Alonso, Gáspár Bakos, Nairn Baliber, Tim Brown, Orlagh Creevey, Jonathan Devor, Ted Dunham, Juan Belmonte, Hans Deeg, Gil Esquerdo, Mark Everett, José Fernández, Scott Gaudi, Márton Hidas, Matt Holman, Luke Kotredes, Géza Kovács, David Latham, Georgi Mandushev, Markus Rabus, Alex Sozzetti, Bob Stefanik, John Trauger, Willie Torres, Russel White, and Josh Winn. Money makes the world go round, and I gratefully acknowledge the financial support of my thesis work with Sleuth from NASA under the grant NNG05GJ29G, issued through the Origins of Solar Systems Program. I wish to recognize and acknowledge the very significant cultural role and reverence that the summit of Mauna Kea has always had within the indigenous Hawaiian community. I am most fortunate to have the opportunity to conduct observations from this mountain. Part of this work is based on observations made with the Spitzer Space Telescope, which is operated by the Jet Propulsion Laboratory, California Institute of Technology under a contract with NASA. This research has made use of the SIMBAD database, operated at CDS, Strasbourg, France, and NASA s Astrophysics Data System Bibliographic Services. This publication also utilizes data products from the Two Micron All Sky Survey, which is a joint project of the University of Massachusetts and the Infrared Processing and Analysis Center/California Institute of Technology, funded by the National Aeronautics and Space Administration and the National Science Foundation. This research has made use of the USNOFS Image and Catalogue Archive operated by the United States Naval Observatory, Flagstaff Station. My thanks to Alicia, Chao, Luke, Sean, and Joanna, my classmates. You helped me survive first year Amigo s margaritas will never taste as good as they did as I drank them with my fellow sufferers! And of course Cathy, Elina, Laura, Margaret, Melissa, Milan and Stuartt, who continued the great tradition of getting to know the First Years, though some more than others. One s fellow graduate students are a tremendous source of advice regarding all aspects of life. My thanks to all the Caltech

vi Astronomy students that I have known over the years. A special mention must be made of Micol and Chin for their help choosing the ring, and of Brian and Danielle for their help choosing our wine! My thanks also to all the Astronomy staff that made my thesis possible, especially to Patrick Shopbell, Anu Mahabul, and Cheryl Southard for keeping my computers working! During my thesis, I visited David Charbonneau quite often at the Harvard Smithsonian Center for Astrophysics. I thank all of the students there for their warm welcome, in particular Cullen, Heather, Jonathan, José, Lisa, and Manuel. I also thank Blue and her family for giving me a second home away from home. Surprisingly, I did find time to meet people outside of Caltech. Karen, our Good Samaritan, has helped me and Cathy through all of our marathon struggles I ll take running the LA marathon over finishing a PhD any day! My quest to complete 26.2 miles of hell was also aided by the presence of the never-say-die Pasadena Pacers. I treasure my time over the last five years with my extended family, especially the new additions, and my friends in Cork. Though separated from me by 6,000 miles, they have continued to be an important part of my life. They have always made time for me during my visits home, and made me feel like I had never left. I would never have attempted the PhD program at Caltech without the support of my parents, Tom and Vera, in particular their relentless faith that I can do anything that I put my mind to. They have given so much to me that can never be repaid, and I am proud to call them not only my parents, but two of my best friends. I know they will always welcome me back in Cork, especially if I fix the computer one more time! It s great having a older sibling. She is always an inspiration, a listener, a friend. I thank my big sister, my Best Sister, Bridget, for always being 18 months older than me, and for letting her younger brother boss her around sometimes. While I struggled to complete my thesis, she managed to finish hers while working a full-time job. She makes it all look easy. And of course, my loving thanks to Cathy, who has gotten to know the real me this year. While preparing for her own defense on the same day as mine, she always

vii kept some energy to bolster me when things got too much for me, and richly deserves being mentioned three times in these acknowledgments. She knows me so well, the only time I managed to surprise her was the day I asked her to marry me. Now that our graduate student days are over, I look forward to a new life together, her hand in mine.

viii Abstract I present the discovery of three transiting planets (TrES-2, TrES-3, and TrES-4) of nearby bright stars made with the ten-centimeter telescope Sleuth as part of the Trans-atlantic Exoplanet Survey (TrES). TrES-2 is the first transiting exoplanet detected in the field of view of NASA s Kepler mission. Of the 20 known transiting exoplanets, TrES-3 has the second shortest period, facilitating the study of orbital decay and atmospheric evaporation. Its visible/infrared brightness makes TrES-3 an ideal target for observations to determine the atmospheric composition. TrES-4 has the largest radius and lowest density of the known transiting planets. These three planets have radii larger than that of Jupiter, and the radius of TrES-4 significantly exceeds predictions from models of hot Jupiters, indicating a possible lack of an energy source in these models. I present the results of Spitzer observations of TrES-2. I reject tidal dissipation of eccentricity as an explanation for the inflated radius, and examine the spectrum for evidence of atmospheric absorption. I have monitored 19 fields each containing 6,000 36,000 stars for evidence of transits. I discuss the rejection of six of my candidate transiting systems from an early field that represent examples of the 67 astrophysical false positives that I encountered in Sleuth data. These six false positives highlight the benefit of a multisite survey such as TrES, and also of comprehensive follow-up of transit candidates. As a further example, I present the candidate GSC 03885-00829 from Sleuth data that was revealed to be a blend of a bright F dwarf and a fainter K-dwarf eclipsing binary. This candidate proved nontrivial to reject, requiring multicolor follow-up photometry to produce evidence of the true binary nature of this candidate.

ix The yield of planets from transit surveys is not yet well constrained or understood. There are numerous factors that affect the predictions such as the amount of correlated photometric noise present in the data. Here I present an analysis of my ability to recover fake transits in TrES data. I examined both the automated transitsearch algorithm and my own visual identification process. I find the recovery rate of my visual analysis to be 87% for those transit candidates that had a sufficiently high signal-to-noise ratio to be flagged by my transit-search algorithm and readily identifiable by eye.

x Contents Acknowledgements Abstract List of Figures List of Tables iv viii xiii xv 1 Transiting Exoplanets 1 1.1 The Search for Other Worlds Using Starlight.............. 1 1.2 Methods for Detecting Hot Jupiters and Other Exoplanets...... 2 1.2.1 Radial Velocities......................... 2 1.2.2 Transits.............................. 7 1.3 Formation, Structure, and Composition of Highly-Irradiated Gas Giants 10 1.3.1 The Birth of Giants........................ 11 1.3.2 A Core or Not a Core, That is the Question.......... 11 1.3.3 Extrasolar Atmospheres..................... 13 1.4 Thesis Motivation Past and Present.................. 14 1.5 Thesis Outline............................... 15 2 Outcome of Six Candidate Transiting Planets from a TrES Field in Andromeda 19 Abstract..................................... 19 2.1 Finding a Needle in a Haystack..................... 20 2.2 Follow-up Observations of Planetary Candidates: A Review..... 23

xi 2.3 Initial Observations of a Field in Andromeda.............. 28 2.4 Searching for Transit Candidates in Andromeda............ 28 2.5 Follow-up of Candidate Transiting Planets............... 34 2.6 Rejecting False-Positive Detections................... 38 2.7 Maximizing the Yield from TrES.................... 45 3 Rejecting Astrophysical False Positives from the TrES Transiting Planet Survey: The Example of GSC 03885-00829 47 Abstract..................................... 47 3.1 Transits versus Blended Eclipses..................... 48 3.2 TrES Telescope Observations of GSC 03885-00829............................. 52 3.3 Spectroscopic Follow-up of GSC 03885-00829.............. 56 3.4 Photometric Follow-up of GSC 03885-00829.............. 59 3.5 Blend Analysis of Observations of GSC 03885-00829............................. 62 3.6 Confirmation of Blend Model...................... 66 3.7 The Necessity of Blend Identification.................. 70 4 TrES-2: The First Transiting Planet in the Kepler Field 72 Abstract..................................... 72 4.1 The Search for Transiting Exoplanets.................. 72 4.2 Observations and Analysis of TrES-2.................. 73 4.3 Estimates of Planet Parameters and Conclusions............ 81 5 TrES-3: A Nearby, Massive, Transiting Hot Jupiter in a 31-Hour Orbit 84 Abstract..................................... 84 5.1 Transiting Exoplanets with Very-Short Orbital Periods........ 85 5.2 Observations and Analysis of TrES-3.................. 86 5.3 Estimates of Planet Parameters and Conclusions............ 92

xii 6 Detection of Planetary Emission from TrES-2 using Spitzer/IRAC 96 Abstract..................................... 96 6.1 Spitzer Observations of the Known Transiting Exoplanets...... 97 6.2 IRAC Observations of TrES-2...................... 98 6.3 Deriving and Modeling Light Curves of TrES-2............. 99 6.4 Atmospheric Models for TrES-2..................... 102 6.5 Searching for Evidence of Atmospheric Absorption........... 103 7 Identifying Transits in TrES Data Sets: The Human Factor 105 7.1 Understanding the Yield of Transit Surveys.............. 105 7.2 Injecting Model Light Curves into a TrES Data Set.......... 107 7.3 BLS and Visual-Recovery Methods of Injected Transits........ 120 7.4 BLS and Visual-Recovery Rates..................... 121 7.5 Discussion................................. 134 8 Summary 135 A TrES-4: A Transiting Hot Jupiter of Very-Low Density 137 Abstract..................................... 137 A.1 Understanding the Mass Radius Relations of Exoplanets....... 138 A.2 Photometry and Spectroscopy of TrES-4................ 138 A.3 Properties of TrES-4 and Discussion................... 145 B Properties of Known Transiting Systems 148 C The Box-fitting Least-Squares (BLS) Transit-Search Algorithm 151 C.1 A General Description of the BLS Algorithm.............. 151 C.2 The Signal Residue and Signal Detection Efficiency.......... 152 Bibliography 154

xiii List of Figures 1.1 Orbital period distribution of the 232 known exoplanets........ 4 1.2 Orbital eccentricity distribution of known exoplanets.......... 6 1.3 Mass-radius and mass-period relations for transiting planets...... 8 2.1 Fraction of potential transit signals identifiable from And0 data.... 27 2.2 Variation in rms residual with magnitude for And0 data........ 31 2.3 TrES light curves of the six candidates from the And0 field....... 32 2.4 Sample spectra of the And0 transit candidates.............. 40 2.5 Radial velocity orbit of T-And0-02022, an F+M eclipsing binary.... 41 2.6 Follow-up photometry of T-And0-03874, a blended eclipsing binary.. 43 2.7 TrES photometry of T-And0-03874 and the neighboring binary.... 44 3.1 TrES light curve of GSC 03885-00829, a blended eclipsing binary... 54 3.2 Location of blend model components on isochrones........... 57 3.3 Sample spectrum of the F-dwarf GSC 03885-00829........... 58 3.4 Color dependence of recovered transits of GSC 03885-00829...... 60 3.5 g-band observations modeled as an F/K+M blend............ 63 3.6 r-band photometry modeled as an F/K+K blend............ 64 3.7 Near-infrared colors for GSC 03885-00829 and an F star........ 67 3.8 Spectrum showing evidence of light from late-type star......... 69 4.1 Relative flux versus time for the TrES-2 transiting system....... 75 4.2 Radial-velocity observations of TrES-2.................. 79 5.1 Radial-velocity observations of TrES-3.................. 90

xiv 5.2 Relative flux versus time for the TrES-3 transiting system....... 93 6.1 Near-infrared relative fluxes from TrES-2 from Spitzer observations.. 100 6.2 Near-infrared contrast ratios for TrES-2................. 102 7.1 Histogram of 2MASS J K s for TrES Lyr1 field............ 107 7.2 Randomized stellar parameters for injected transits........... 108 7.3 Randomized planetary parameters for injected transits......... 109 7.4 Randomized transit impact parameters and orbital inclinations.... 111 7.5 Randomized values for first two BLS parameters............ 113 7.6 Randomized values for final two BLS parameters............ 114 7.7 Example of model transits......................... 115 7.8 Example of TrES light curve with injected transits........... 116 7.9 Histograms of BLS SDEs for fake data set................ 118 7.10 Example of an injected transit candidate with a high SDE....... 119 7.11 SDE and V T versus period for BLS-recovered transits.......... 122 7.12 Histogram of SDEs for BLS-recovered transit candidates........ 123 7.13 Ratio of to σ versus period for BLS-recovered transits........ 124 7.14 BLS recovery rates for different orbital periods and planetary radii.. 126 7.15 BLS recovery rates for different stellar radii and impact parameters.. 127 7.16 SDE and V T versus period for visually recovered transits........ 128 7.17 Histogram of SDEs for visually recovered transit candidates...... 129 7.18 /σ versus period for visually recovered transits............ 130 7.19 Visual-recovery rates for different periods and planetary radii..... 132 7.20 Visual recovery of different stellar radii and impact parameters.... 133 A.1 High-precision follow-up z-band and B-band photometry of TrES-4.. 142 A.2 Keck/HIRES radial velocity observations of TrES-4........... 144

xv List of Tables 2.1 TrES labels, 2MASS and GSC designations, and approximate V magnitudes for And0 candidate transiting systems.............. 29 2.2 Transit properties for the six TrES And0 candidates.......... 30 2.3 Photometric and spectroscopic properties of the six TrES And0 candidates 35 2.4 Details of spectroscopic observations of the six TrES And0 candidates 36 3.1 Data for GSC 03885-00829........................ 55 4.1 System parameters for TrES-2 planet.................. 74 4.2 System parameters for TrES-2 parent star................ 76 4.3 Relative radial-velocity measurements of TrES-2............ 78 5.1 TrES-3 parent star............................. 88 5.2 Relative radial-velocity measurements of TrES-3............ 89 5.3 TrES-3 planet............................... 90 A.1 TrES-4 host star.............................. 141 A.2 TrES-4 planet parameters......................... 143 A.3 Radial-velocity measurements of TrES-4................. 145 B.1 Properties of the known transiting planets................ 152 B.2 Properties of the known transiting-planet host stars.......... 153

1 Chapter 1 Transiting Exoplanets 1.1 The Search for Other Worlds Using Starlight Astronomy in the 21st century has seen significant progress made in the age-old question of our uniqueness in the universe. Less than two decades since the first discoveries of planetary-mass systems around stars other than our Sun (Latham et al., 1989; Wolszczan & Frail, 1992; Mayor & Queloz, 1995), we now know of 200 planetary systems, and the rate of new discoveries is increasing. The planets in these new systems are known as extrasolar planets or exoplanets for short. Since the reflected or emitted light from a planet is very dim compared with starlight, the vast majority of detections have been made by studying the light from the host star as the star interacts with the planet. Software and hardware improvements have made possible the detection of minute changes in observations of these stars. This thesis work, and the study of exoplanets in general, utilizes telescopes both on the ground and in space, and ranging from ten-centimeter to ten-meter in diameter. With these tools, we are able to test our understanding of how gas giants are formed and how their internal structure and chemical composition vary with environment. With each new discovery, we make another step toward the detection of a solar system analog or a mirror Earth in the habitable zone of a star. Before continuing, it should be noted that a distinction is made by the International Astronomical Union (IAU), if not by nature between different types of stellar companions that are thereby segregated by mass. The evolution of celestial objects

2 that fuse hydrogen in their cores is quite different from the evolution of objects without this fusion. Hence, only the hydrogen-burning bodies (those that have at least 80 times the mass, M Jup, of Jupiter) are called stars. Objects that are less massive than stars, yet can fuse deuterium in their cores, are known as brown dwarfs, and have masses greater than 13M Jup. Finally, the least massive stellar companions are planets. The latter are the target of my survey, and the planets found during this thesis work have masses m p M Jup, well below the 13M Jup limit, hence different interpretations of the mass boundaries will not affect my results. 1.2 Methods for Detecting Hot Jupiters and Other Exoplanets The majority of the known exoplanetary systems contain planets similar in mass to Jupiter and the other solar system giant planets, rather than the lighter terrestrial planets such as Earth. This is because most of the exoplanets were found by measuring a change in the motion of their star caused by the gravitational presence of the orbiting planet, a method that is inherently more sensitive to massive planets. The velocity of a star in space can be split into two components: the radial velocity along the line of sight to the star (either toward or away from us), and the tangential velocity across the sky, perpendicular to the line of sight. Although there are several other successful methods to find exoplanets, in this thesis I will concentrate on a planet search that uses the measurement of radial-velocity variations to confirm the existence of planets observed to transit across stellar disks. 1.2.1 Radial Velocities The radial velocity of a star is determined from the Doppler shifts of the stellar spectral lines due to the motion of the star. By measuring the variation with time of the radial velocities of the star (of mass M ) and deriving the stellar orbit around the barycenter of the system, we can determine several orbital parameters of the planet,

3 such as the orbital period P and the eccentricity e. We can also estimate its mass, or at least a lower limit to the mass, from these orbital parameters and the amplitude of the variation. The product of the planetary mass m p and the sine of the orbital inclination i can be computed from the mass function: (m p sin i) 3 (M + m p ) 2 = PK3 2πG (1 e2 ) 3/2, where K is the semi-amplitude of the radial velocity variation and G is the universal gravitational constant. This product is known as the minimum mass of the planet. Assuming a circular orbit and that m p << M, we can write this as: ( ) ( ) 1/3 ( ) 2/3 K P M m p sin i 1M Jup. (1.1) 200 m s 1 day M Thus, the gravitational effect of Jupiter on the Sun is quite small, resulting in a motion of 12.5 m s 1 or 45 km hr 1, whereas a 1M Jup planet in a one-day period around the sun would induce a motion of K = 200 m s 1. The ability to measure from many parsecs away these small orbital accelerations requires measuring shifts in the stellar spectral lines with a precision of roughly one thousandth of the width of a spectral line. Hardware and software developments have improved this precision to the point where it is possible to detect planets with minimum masses of 5M ( 0.02M Jup ; e.g., Gliese 581c; Udry et al., 2007). Although the radial-velocity surveys are the most prolific planet campaigns, they suffer from several limitations. The target stars are observed one at a time, requiring a heavy investment of large telescope time. The surveys are restricted to the brightest stars in the sky, again due to limited telescope resources. Finally, the precision of these measurements is not likely to be indefinitely improved, due to an astrophysical limit. Stars are asteroseismically active, producing radial-velocity jitter on the scale of 1 10 m s 1 (Saar, Butler, & Marcy, 1998; Santos et al., 2000; Wright, 2005). It is thought that this jitter will hinder achieving precisions much better than currently obtained.

4 Figure 1.1 Orbital period distribution of the 232 known exoplanets as a function of their minimum masses. Jupiter is shown for comparison as a red, filled circle. Relatively few massive planets with short orbital periods have been found, despite the bias for such planets in the radial-velocity surveys. An approximate detection threshold (defined as K = 4σ, see text) is shown as a dotted line.

5 The enlarging catalog of planets from radial-velocity surveys of stars in the solar neighborhood facilitates studying the statistics of exoplanets. For example, the frequency of planets can be estimated from the yield of these surveys, and is currently estimated at 5% for planets of 0.5 8M Jup within 3 AU (Udry et al., 2007). There is a correlation between the metallicity of stars and this frequency of planets around them (Gonzalez, 1997; Fischer & Valenti, 2003; Santos, Israelian, & Mayor, 2004; see Gonzalez, 2006 for a review). We can also see that known exoplanets have a wide range in orbital periods, as shown in figure 1.1. This figure also shows a distinct decreasing trend in the mass of the planet with the short orbital period (Zucker & Mazeh, 2002; Udry, Mayor, & Santos, 2003), which may be related to the way in which these gas giants come to be located so close to their stars (see section 1.3.1). The lack of low mass planets at longer periods can be explained in terms of the detection threshold of radial-velocity surveys. Planets can be detected if they induce a motion K > 4σ in their star, where σ is the uncertainty in each velocity measurement (Marcy, Cochran, & Mayor, 2000). Recent discoveries are made with uncertainties as low as σ = 1m s 1 (Lovis et al., 2006). The associated detection threshold, calculated using equation 1.1 (assuming R = R ), is shown in figure 1.1. Many of the known extrasolar gas giants have orbital periods much less than that of Jupiter (again, see figure 1.1), much to the surprise of the early discoverers. Extrasolar gas giants that orbit their stars within 0.1 AU are known as hot Jupiters, referring to the intense insolation experienced in a relative orbit well inside that of Mercury. Tidal interactions (see, e.g., Rasio & Ford, 1996) with the nearby star result in circular orbits for most of these close-in planets (see figure 1.2). Several planet searches such as this thesis work concentrate on these objects as they are both easier to find (due to their rapidly repeating signals) and fascinating to study (due to their extreme environment close to the star). The proximity to the star also increases the probability that they will be observed to pass across the stellar disk. The radial-velocity method achieves its full potential when combined with such transit observations, as this combined method allows a precise estimate of the mass and radius of the planet.

6 Figure 1.2 Orbital eccentricities of known exoplanets as a function of their orbital distance. Jupiter is shown for comparison as a red, filled circle. Most of the hot Jupiters (those to the left of the dotted line) have negligible eccentricity as expected from tidal circularization.

1.2.2 Transits 7 For every system of gravitationally bound celestial objects, there is a probability that we will observe these objects passing in front of each other, eclipsing the light from the other. In the case of a hot Jupiter with a radius r p in a circular orbit around a star of radius R, this probability is given by: P = r p + R, a R a, 10% ( R R ) ( a ) 1, 0.05AU where 0.05 AU is the median orbital distance for systems containing hot Jupiters (Sackett, 1999). A planet in an 0.05 AU orbit around a solar-sized star (with P 10%) is therefore an ideal candidate to monitor for the transit of the planet across the star. The passage of the planet across the stellar disk will block light from the star and reduce the flux by F. For example, a transit of the Sun by Jupiter would block 1% of the sunlight. Assuming that R can be determined from spectroscopy, the radius of the planet can then be calculated from: r p = R F. Such a transit would have a duration D of : ( D = P π arcsin R 1 + r ) 2 p b a sin i 2, R where b (= a cos i/r ) is the impact parameter (Seager & Mallén-Ornelas, 2003). For a Jupiter-sized planet that transits the equator of a solar-type star with a 4-day period, the transit duration D 2.5 hours. In order for a transit to occur, the orbital inclination must be close to 90, hence

8 Figure 1.3 Mass-radius (top) and mass-period (bottom) relations for the transiting planets known at the time of writing. The dot-dashed lines represent different densities. There is a spread of radii for planets of the same mass, and reproducing the largest radii is still beyond current structural models. There also appears to be a decreasing trend in mass with period in the bottom panel. In the 1 2-day period range, the planets all have m p M Jup. This may indicate different formation mechanisms for these planetary systems than those at longer periods.

9 the minimum mass derived for a transiting planet using equation 1.1 is a close approximation of the true planetary mass. By combining observations of transits and radial-velocity variations, astronomers have obtained mass and radius precise estimates of 20 transiting exoplanets (at the time of writing; see figure 1.3) with which to test structural models for these giant balls of gas. (For reference, I have placed tables of the properties of these transiting systems in appendix B.) The masses of the transiting planets are clustered around 1M Jup. However, the discovery of the least massive (GJ 436b) and most massive (HAT-P-2b) transiting planets to date suggest that over time the mass distribution will resemble more that of the radial-velocity planets (see figure 1.1). figure 1.3 also shows the variation in orbital period with mass. Although results thus far are limited by small number statistics, it appears that the more massive transiting planets have shorter orbital periods, and every m p < 1M Jup planet has an orbital period greater than 2 days. This trend, which is in the opposite direction to the corresponding trend for the known radial-velocity planets (see figure 1.1), was first noted by Mazeh, Zucker, & Pont (2005) and Gaudi, Seager, & Mallen-Ornelas (2005). If the trend continues to be observed as more transiting planets are discovered in the range of orbital periods 1 5 days, it may indicate a different origin for the planets with one-day periods, perhaps by tidal capture rather than inward migration (Gaudi & Winn, 2007). In contrast to the number of known eccentric radial-velocity planets at small orbital separations (see figure 1.2), there are only two known transiting planets (GJ 436b and HAT-P-2b) in this category. Again, over time when more transiting systems are known, it is likely that this number will significantly increase. As well as providing the most precise masses and radii known for extrasolar planets, nearby transiting planets present some other unique opportunities to study their nature: 1. Space-based telescopes offer the promise of exquisitely precise transit light curve with which to look for evidence of rings and satellites orbiting the gas giant (see, e.g., Brown et al., 2001).

10 2. Radial-velocity observations of the star during transit can measure the strength of the Rossiter-McLaughlin effect (McLaughlin, 1924; Rossiter, 1924), an apparent radial-velocity variation caused by the blocking of light from the differentially rotating star. The strength of the variation indicates the misalignment of the orbital and rotation axes of the system. 3. We can also observe the starlight as it transmits through the planetary atmosphere during transit, and look for spectral features indicative of the chemical composition of the planet (Charbonneau et al., 2002; Vidal-Madjar et al., 2003, 2004; Deming et al., 2005a; Barman, 2007). 4. The relative flux of a hot Jupiter to that of the star is greatest in the infrared. Since a transiting system undergoes a secondary eclipse as well as a primary eclipse (the transit), we can estimate the strength of this emission by comparing the system flux before the secondary eclipse, and the flux during the secondary eclipse when the star blocks the planetary flux. Infrared emission has already been observed from several transiting planets using Spitzer (see, e.g., Charbonneau et al., 2005; Deming et al., 2005b), and recently the first spectra of exoplanets were obtained (Richardson et al., 2007; Swain et al., 2007; Grillmair et al., 2007), as well as the first map of the emission variation across a planet (Knutson et al., 2007a). 1.3 Formation, Structure, and Composition of Highly-Irradiated Gas Giants Transiting systems provide precise measurements of planetary masses and radii and facilitate the direct study of the orbital alignment and spectral features of the planet. In this section, I will explain how these observations serve as important constraints for models of the initial formation, internal structure and atmospheric composition of these distant gas giants. Before the discovery of these giants in such extreme environments, the only constraints of theoretical models were the solar system gas

11 giants. Models based on our own planetary system proved to be inadequate for hot Jupiters. 1.3.1 The Birth of Giants The canonical model for the formation of gas giants prior to the discovery of hot Jupiters nicely reproduced the giant planets of our solar system. In this core accretion model (see, e.g., Pollack, 1984; Pollack et al., 1996), the gas giant begins as a protoplanetary core within a nebular disk and grows through collisions with other protoplanets and the accretion of gas from the surrounding disk. When the core is massive enough, a period of runaway gas accretion occurs, and a giant planet is born. This core formation relies on a suitable quantity of icy particles in the surroundings to form planetesimals. With the discovery of 51 Peg b (Mayor & Queloz, 1995), a giant planet close to its star and well inside the ice line (the boundary beyond where most material is in solid, rather than gaseous, form), the validity of this theory was called into question: how could this planet have formed a solid core so close to the star? The most likely explanation is that hot Jupiters like 51 Peg b formed outside the ice line (as did the solar system gas giants), and then moved or migrated inward toward the star (Goldreich & Tremaine, 1980; Lin, Bodenheimer, & Richardson, 1996; Trilling et al., 1998). This migration is enabled by gravitational interactions between the planet and the disk. Whether the migration is halted by some mechanism or the observed planet survived simply because the disk dissipated in time to prevent further migration is still unresolved. 1.3.2 A Core or Not a Core, That is the Question An alternative formation model for gas giants was proposed by Boss (1997). In his gravitational instability model, gas giants can be formed rapidly as collapsing instabilities in the nebula. A consequence of this rapid formation is that the gas giant is thought not to have a substantial core, although some heavy elements may be accumulated through planetesimal bombardment.

12 The ability to measure the masses and radii of hot Jupiters allows us to test for the presence or absence of cores, and thus differentiate between the two models. Models of gas giants that include a substantial core have a smaller radius for the planet, and this effect is larger for the less massive giants. There is some evidence for core accretion in the observations of transiting planets. HD 149026b is a transiting hot Saturn whose radius is so small for its mass that a large core of approximately 70M of heavy elements is implied by the models (Sato et al., 2005; Charbonneau et al., 2006; Fortney et al., 2006b). However, ever since the discovery of HD 209458b, we have struggled to adapt these models to explain every extrasolar mass-radius relation. We have now identified several transiting gas giants whose radii exceed our predictions for their masses; the values for the remaining planets are in agreement with models either with or without a core (see, Laughlin et al., 2005; Charbonneau et al., 2007a, for a review). In order to explain these inflated planets, several additional energy sources were proposed that could slow the contraction of the planet after formation, and would result in a larger radius. Bodenheimer, Lin, & Mardling (2001) and Bodenheimer, Laughlin, & Lin (2003) suggested the presence of an additional but unseen planet in the transiting system that would continuously pump the orbital eccentricity of the detected planet. The tidal circularization of this nonzero eccentricity would produce the energy internal to the planet required to maintain the large radius. Winn & Holman (2005) proposed a similar tidal dissipation, this time of a nonzero obliquity of a planet in a Cassini state. The explanation put forth by Guillot & Showman (2002) and Showman & Guillot (2002) was that some of the intense insolation creates atmospheric winds and thereby contributes thermal energy to the interior. Burrows et al. (2007) explored the effect of substantially increased planetary metallicity on the opacity of the planetary atmosphere and suggested that the resultant increased opacity would act to slow the heat loss from and hence contraction of the planet. Burrows, Sudarsky, & Hubbard (2003) and Baraffe et al. (2003) pointed out an effect due to the difference between the observed planetary radius (corresponding to an optical depth of unity at some wavelength) and the theoretical

13 radius (at the 1-bar level) that could account for 5% of the discrepancy between the two. However, no satisfying solution has been found to date. The kinetic energy source of Guillot & Showman (2002) should apply to every hot Jupiter, yet HD 209458b and TrES-1 have similar masses but significantly different radii. Deming et al. (2005b) refuted the possibility that tidal damping of a nonzero eccentricity was an energy source for the inflated planet HD 209458b by deriving a negligible eccentricity from observations of a secondary eclipse. Also, Fabrycky, Johnson, & Goodman (2007) has rejected obliquity tides as not providing enough energy through dissipation to account for large radii of hot Jupiters. 1.3.3 Extrasolar Atmospheres Although giant planets consist mainly of gas, namely molecular hydrogen and helium, both the absorption rate of incident radiation, and the emission from (and hence cooling rate of) the planet is determined mostly by a thin outer layer of these gaseous molecules. This thin layer is called the atmosphere of the planet. The composition of the atmosphere, and the gas giant itself, is expected to be roughly the composition of the nebula from which it formed, although substantially enriched due to bombardment of planetesimals. The presence or absence of specific chemical species in the atmosphere is then dependent on the temperature and pressure gradients of the planet. The metallicity of the atmosphere can be a significant effect. The infrared color of TrES-1 (Charbonneau et al., 2005) is too red when compared to model spectra and other planets, and Fortney et al. (2005) explained this as due to a metallicity 3 5 times solar. Once again, the extreme nature of the environment of hot Jupiters presents a new challenge to models of gas giants, this time of their atmospheres. Jupiter has an effective temperature of 125 K, has an intrinsic luminosity due to ongoing contraction roughly equal to its luminosity due to reradiated solar flux, and almost completely redistributes the thermal energy from this flux from the dayside to the

14 night-side of the planet. In contrast, hot Jupiters have equilibrium temperatures of up to 1500 K. Of their luminosity, 99.99% is due to the re-radiation of the insolation (Marley et al., 2007). This intense radiation may result in hot substellar spots, and could cause day-night temperature differences of up to 500 K, and winds of up to 2 km s 1 (Showman & Guillot, 2002). The efficiency of the redistribution of the heat from this spot may vary from planet to planet. Due to the high effective temperatures, atmospheric models predict different dominant molecular absorbers in the spectra of the hot Jupiters than of the solar system giants. At the visible wavelengths, Na and K produce strong absorption features, whereas in the infrared, H 2 O, CO, CH 4, and CH 3 are all noticeable absorbers. The exact planetary spectrum depends on the amount of insolation absorbed by the planet, the redistribution of the resultant thermal energy, and the presence or absence of high clouds of condensates. Unlike the solar system giants for which we have direct estimates of the solar flux and of the chemical compositions, we can be certain only of the incident radiation onto these planets. However, if we measure the amount of flux from the night-side of a transiting planet (during a transit) and from the day-side (during a secondary eclipse), we can deduce how much energy is being transported from the dayside, and hence measure the efficiency of this redistribution. Understanding the atmospheric makeup of the hot Jupiters may provide us with an explanation for the inflated exoplanets, as the predictions for these radii are dependent on understanding the cooling rates of the planets. 1.4 Thesis Motivation Past and Present At the beginning of my thesis work, only one transiting planet, HD 209458b, was known. The radius of this planet was larger than could be explained by extrapolating models of solar system gas giants to account for intense insolation. This made the discovery of additional transiting planets very important to allow determination of whether this planet was merely an anomaly. Transit surveys also held the promise of being the most effective way to obtain precise planetary masses and radii to provide

15 constraints for models. In contrast to the radial-velocity surveys of individual stars, transit surveys monitor thousands of stars at a time for evidence of a planetary companion, although the expected frequency of detections was not well understood. With the TrES team s discovery of TrES-1 (Alonso et al., 2004a), wide-field surveys proved our ability to find transiting planets around stars bright enough to provide these desired precise constraints. It was clear that my wide-field transit survey could make a substantial contribution even by discovering only one or two transiting planets. By extending the coverage of the parameter space for exoplanets, I could hope to improve the understanding of the wide range of radii for planets of similar masses. The bright stars that I would find as part of my survey would also be ideal targets for detailed studies of the atmosphere and neighboring objects with space-based telescopes. The goals that motivate my thesis work still hold today, since the number of known transiting planets (19) is still quite low. The planets found by TrES and other surveys continue to be much sought after, and quickly observed with the best instruments available to uncover the true nature of each planet. 1.5 Thesis Outline The goals of my thesis were to both detect and explore new transiting planets. Much of my thesis work involved participation in the Trans-atlantic Exoplanet Survey (TrES), a network of three ten-centimeter optical telescopes used to search for nearby transiting planets. For four years, I have operated and maintained the telescope Sleuth at Palomar Observatory, obtaining observations of TrES fields almost every night (weather permitting). I have reduced these observations, and searched the resulting time series of field stars for transit signals. For each transit candidate that I identified, it has been my responsibility to then coordinate follow-up spectroscopic observations, obtained by D. Latham and collaborators, and follow-up photometric observations, obtained by me and several TrES collaborators. I pursued my own follow-up photometry us-

16 ing Sherlock (Kotredes et al., 2004) or the Palomar 1.5-meter telescope (see, e.g., Holman et al., 2006, for an example of the 1.5-meter photometry), and performed modeling of this photometry to derive precise transit parameters. Based on the results of each new set of observations, I removed false positives from my candidate list. Twice during my thesis, I obtained high-dispersion observations of my most promising transit candidates with Keck/HIRES, aided by D. Charbonneau, G. Torres, and A. Sozzetti, who performed the rapid data analysis needed to review the data while still at Keck. The subsequent required blend analysis and photometric modeling of the three planets that I discovered were performed by G. Torres and D. Charbonneau, respectively. Having discovered TrES-2, I then turned my attention to analyzing the Spitzer observations of this planet, obtained by my collaborator J. Harrington. The theoretical interpretation of the data was carried out in collaboration with S. Seager. Most of the chapters in this thesis are based on material published during my time at Caltech. I have presented them in a logical order, rather than ordered by publication date. In chapter 2, I outline how I obtained observations of 26,495 stars in a single field using Sleuth, and my analysis of the data. I present six candidates that I identified from a transit search of the photometry. I explore the methods used by the TrES team to reject astrophysical false postives, and explain how the six stars were culled from our candidate transiting system list. In this chapter, I also demonstrate the advantages of a multisite network such as TrES for obtaining better phase coverage and confirming the reliability of these candidates. Most transit surveys have obtained significant experience with these typical astrophysical false positives. In chapter 3, I discuss an early example of a more insidious false positive, GSC 03885-00829, that I identified as a candidate transiting planet. This system proved to be a solitary star blended with an eclipsing binary. The relative faintness of the binary prevented the detection of its radial-velocity signal. Multicolor photometry revealed a color-dependent eclipse, indicative of the binary

17 nature. In this chapter, I go into more detail about how I produce light curves from Sleuth photometry and select transit candidates. In chapter 4, I present the discovery in one of the Sleuth fields of a massive gas giant TrES-2 that was confirmed from the spectroscopic orbit derived from my observations with Keck/HIRES. TrES-2 is the first transiting planet known to reside in the field of view of the upcoming NASA Kepler mission that will search for new exoplanets. Similarly, in chapter 5, I present TrES-3, another massive transiting planet from a Sleuth field, this time with a very-short orbital period. This discovery has further cemented the hypothesis that the nearby exoplanets discovered by widefield transit surveys and the more distant gas giants found by narrow-field transit surveys have the same intrinsic period distribution. I have also included in appendix A the discovery paper for TrES-4, the third planet found in a Sleuth field. This planet has the largest radius and lowest density of the currently known transiting planets. Although each newly discovered transiting planet automatically provides a new constraint for theoretical structural models for these Jovian planets, the nearby transiting planets also present promises of a more detailed exploration. Spitzer has been a leading source of new information about exoplanets, and in chapter 6, I present my analysis of the first Spitzer/IRAC observations of TrES-2, and determine a negligible orbital eccentricity for this hot Jupiter. There is no evidence of strong atmospheric absorption from these observations, but future analysis including additional Spitzer observations will be more conclusive. Finally, I examine the yield of planets from this survey in comparison with predictions. In chapter 7, I present initial results from a study of my ability to recover fake transit light curves injected into real data from a TrES field. Rather than just concentrate on the recovery rate of the transit-search algorithm I employ, I explore the human element of identifying worthy planet candidates, and the resultant recovery rate of transit candidates. The photometric data from Sleuth have also been of value in the study of eclipsing binaries. As this is not related to my thesis goals, I simply refer the reader to Creevey et al. (2005) and Devor et al. (2007, in preparation) as examples of this use

of Sleuth data. 18

19 Chapter 2 Outcome of Six Candidate Transiting Planets from a TrES Field in Andromeda 1 Abstract Driven by the incomplete understanding of the formation of gas giant extrasolar planets and of their mass-radius relationship, several ground-based, wide-field photometric campaigns are searching the skies for new transiting extrasolar gas giants. As part of the Trans-atlantic Exoplanet Survey (TrES), in 2003/2004 we monitored approximately 30,000 stars (9.5 V 15.5) in a 5.7 5.7 field in Andromeda with three telescopes over five months. We identified six candidate transiting planets from the stellar light curves. From subsequent follow-up observations we rejected each of these as an astrophysical false positive, i.e., a stellar system containing an eclipsing binary, whose light curve mimics that of a Jupiter-sized planet transiting a Sun-like star. We discuss here the procedures followed by the TrES team to reject false positives from our list of candidate transiting hot Jupiters. We present these candidates as early examples of the various types of astrophysical false postives found in the TrES campaign, and discuss what we learned from the analysis. 1 This chapter has been published previously as O Donovan et al. 2007a, ApJ, 662, 658.

20 2.1 Finding a Needle in a Haystack At the time of writing, there are 14 extrasolar planets for which we have measurements of both the planetary radius and mass (see Charbonneau et al., 2007a, for a review; McCullough et al. 2006; O Donovan et al. 2006a, see chapter 4; Bakos et al. 2007; Collier Cameron et al. 2007). These gas giants have been observed to transit their parent stars and have supplied new opportunities to study Jupiter-sized exoplanets, in particular their formation and structure. Studies of the visible and infrared atmospheric spectra are possible for the nine nearby (d < 300 pc) transiting planets (Charbonneau et al., 2002; Vidal-Madjar et al., 2003; Deming et al., 2005a,b; Charbonneau et al., 2005). The incident flux from the close-by ( < 0.05 AU) star on each of these hot Jupiters results in an inflated planetary radius. Current theoretical models that include this stellar insolation can account for the radii of only five of these nine nearby planets. HD 209458b (Charbonneau et al., 2000; Henry et al., 2000), TrES-2 (O Donovan et al., 2006a, see chapter 4), HAT-P-1b (Bakos et al., 2007), and WASP-1b (Collier Cameron et al., 2007) have radii larger than predicted (see Laughlin et al., 2005 and Charbonneau et al. 2007a for reviews of the current structural models for insolated hot Jupiters). The sparse sampling and limited understanding of the mass-radius parameter space for extrasolar planets continue to motivate the search for new transiting planets. There are several small-aperture wide-field surveys targeting these objects, such as the Berlin Exoplanet Search Telescope (BEST; Rauer et al. 2004), the Hungarian Automated Telescope (HAT) network (Bakos et al., 2002, 2004), the Kilodegree Extremely Little Telescope (KELT; Pepper, Gould, & Depoy 2004), the Super Wide Angle Search for Planets (Super- WASP; Street et al. 2003), Vulcan (Borucki et al., 2001), and XO (McCullough et al., 2005), as well as deeper surveys like the Optical Gravitational Lensing Experiment (OGLE-III; Udalski et al. 2002) that is probing the Galactic disk. We are conducting a transit campaign, the Trans-atlantic Exoplanet Survey 2 (TrES), using a network of three ten-centimeter telescopes with a wide longitudi- 2 See http://www.astro.caltech.edu/~ftod/tres/.

21 nal coverage: Sleuth 2 (located at Palomar Observatory, California; O Donovan et al. 2004), the Planet Search Survey Telescope (PSST; Lowell Observatory, Arizona; Dunham et al. 2004), and the STellar Astrophysics and Research on Exoplanets 3 telescope (STARE; on the isle of Tenerife, Spain; Alonso et al. 2004b). Over several months the telescopes monitor a 5.7 5.7 field of view containing thousands of nearby bright stars (9.5 V 15.5), and we examine the light curves of stars with V 14.0 for repeating eclipses with the short-period, small-amplitude signature of a transiting hot Jupiter. We have discovered two transiting planets so far: TrES-1 (Alonso et al., 2004a) and TrES-2 (O Donovan et al., 2006a, see chapter 4). In order to find these two transiting planets we have processed tens of candidates with light curves similar to that of a Sun-like star transited by a Jupiter-sized planet. For a typical TrES field at a Galactic latitude of b 15, we find 10 candidate transiting planets (see, e.g., Dunham et al., 2004). We expect few of these to be true transiting planets, and the remainder to be examples of the various types of astrophysical systems whose light curves can be mistaken for that of a transiting planet (see, e.g., Brown, 2003; Charbonneau et al., 2004). These are: (i) low-mass dwarfs eclipsing high-mass dwarfs, (ii) giant+dwarf eclipsing binaries, and (iii) grazing incidence main-sequence eclipsing binaries, with eclipse depths similar to the 1% transit depth of a hot Jupiter, and (iv) blends, where a faint eclipsing binary and a bright star coincide on the sky or are physically associated, mixing their light, with the observed eclipse depth reduced to that of a transiting planet. We also encounter occasional photometric false positives, where the transit event is caused by instrumental error, rather than a true reduction in flux from the candidate. Brown (2003) estimates the relative frequency of these astrophysical false positives. For 3 See http://www.hao.ucar.edu/public/research/stare/stare.html.

22 a STARE field in Cygnus, he predicts that from every 25,000 stars observed with sufficient photometric precision to detect a transit, one can expect to identify one star with a transiting planetary companion. However, for this field near the Galactic plane (b 3 ), the number of impostor systems identified as candidate planets will outnumber the true detections by an order of magnitude. (The yield of eclipsing systems from such transit surveys depends on the eclipse visibility, which is the fraction of such systems with a given orbital period for which a sufficient number of eclipses could be observed for the system to be detected. This visibility varies with weather conditions during the observations and the longitudinal coverage of the telescopes used.) Of the false positives, approximately half are predicted to be eclipsing binaries and half to be blends. A careful examination of the light curve of a transit candidate may reveal evidence as to the nature of the transiting companion. Seager & Mallén-Ornelas (2003) present an analytic derivation of the system parameters that can be used to rule out obvious stellar systems. If the light curve demonstrates ellipsoidal variability, this indicates the gravitational influence of a stellar companion (Drake, 2003; Sirko & Paczyński, 2003). These tests have been used to great effect on the numerous candidates (177 to date) from the OGLE deep-field survey (Drake, 2003; Sirko & Paczyński, 2003; Pont et al., 2005; Bouchy et al., 2005a), and candidates from wide-field surveys (see, e.g., Hidas et al., 2005; Christian et al., 2006). The initial scientific payoff from each new transiting hot Jupiter comes when an accurate planetary mass and radius have been determined, which can then be used to test models of planetary structure and formation. These determinations require a high-quality light curve together with a spectroscopic orbit for the host star. For each TrES target field we follow a procedure of careful examination of each candidate, with follow-up photometry and spectroscopy to eliminate the majority of false positive detections and obtain a high-quality light curve before committing to the final series of observations with ten-meter class telescopes to determine the radial velocity orbit of the candidate planet. This procedure is similar to those discussed by Charbonneau et al. (2004) and Hidas et al. (2005). Here we discuss our follow-up

23 strategy (section 2.2) and present the step-by-step results of this procedure for a field in Andromeda, one of the first fields observed by all three nodes of the TrES network. We describe the TrES network observations in section 2.3 and outline the initial identification of six candidates from the stellar light curves in section 2.4. Based on our follow-up observations of these candidates (section 2.5), we were able to conclude that each was an astrophysical false positive (section 2.6). 2.2 Follow-up Observations of Planetary Candidates: A Review The light curves from small wide-angle telescopes are not of sufficient quality to derive an accurate radius ratio for the purpose of both false positive rejection and planetary modeling, so high-quality follow-up photometric observations with a larger telescope are needed. Recent experience suggests that a photometric accuracy better than 1 mmag with a time resolution better than one minute can be achieved with a meter-class telescope at a good site, and such observations can deliver radius values good to a few percent and transit times good to 0.2 minutes (Holman et al., 2006, 2007a). For smaller telescopes, scintillation can limit the photometric precision at this cadence (Young, 1967; Dravins et al., 1998). The wide-angle surveys by necessity have broad images, typically with FWHM values of 20. Thus, there is a significant probability of a chance alignment between a relatively bright star and a fainter eclipsing binary that just happens to be nearby on the sky. Photometric observations with high spatial resolution on a larger telescope can be used to sort out such cases by resolving the eclipsing binary (see, e.g., Charbonneau et al., 2004). In some instances these systems can also be detected using the wide-angle discovery data, by showing that there is differential image motion during the transit events, even though the eclipsing binary is unresolved. Photometry can also be successful in identifying a triple. If the color of the eclipsing binary is different enough from that of the third star, high-quality multicolor light curves

24 will reveal the color-dependent eclipse depths indicative of such a system (see, e.g., Tingley, 2004; O Donovan et al., 2006b, see chapter 3). A practical problem for this follow-up photometry is that the transiting-planet candidates do not emerge from the wide-angle surveys until late in the observing season, when the observability of the candidates do not permit full coverage of a transit event. With only partial coverage of an event it is difficult to remove systematic drifts across the event, reducing the accuracy of the derived transit depth. Furthermore, full coverage of a transit is important for deriving very accurate transit timings. Without accurate ephemerides, the error in the predicted transit times during the next observing season may be too large to facilitate follow-up photometric observations. The typical duty cycle for a transit is a few hours over a period of a few days, i.e., a few percent. Therefore, if the follow-up photometry does not confirm a transit, the interpretation is ambiguous. The ephemeris may have been too inaccurate, or perhaps the original transit event was a photometric false detection. One approach to recovering transits and providing an updated ephemeris for highquality photometric observations with a larger telescope is to monitor candidates with intermediate sized telescopes, such as TopHAT in the case of the HAT survey (Bakos et al., 2004), Sherlock (Kotredes et al., 2004) in the case of TrES, or teams of amateur telescopes (McCullough et al., 2006) in the case of XO. A second approach to confirming that a candidate is actually a planet is to obtain very precise radial velocities to see whether the host star undergoes a small reflex motion as expected for a planetary companion. This approach has the advantage that the velocity of the host star varies continuously throughout the orbit, so the observations can be made at any time with only modest attention to the phasing compared to the photometric period. The ephemeris can then be updated using the velocities, to provide reliable transit predictions for the follow-up photometry. A second advantage is that a spectroscopic orbit is needed anyway to derive the mass of any candidate that proves to be a planet. The big disadvantage of this approach is that a velocity precision on the order of 10 m s 1 is needed, which requires access to a large telescope with an appropriate spectrograph.

25 For the follow-up of transiting-planet candidates identified by TrES, we have adopted a strategy designed to eliminate the vast majority of astrophysical false positives with an initial spectroscopic reconnaissance using the Harvard Smithsonian Center for Astrophysics (CfA) Digital Speedometers (Latham, 1992) on the 1.5-meter Wyeth Reflector at the Oak Ridge Observatory in Harvard, Massachusetts and on the 1.5-meter Tillinghast Reflector at the F. L. Whipple Observatory (FLWO) on Mount Hopkins, Arizona. We aim to observe candidates spectroscopically during the same season as the discovery photometry. These instruments provide radial velocities good to better than 1 km s 1 for stars later than spectral type A that are not rotating too rapidly, and thus can detect motion due to stellar companions with just two or three exposures (see, e.g., Latham, 2003; Charbonneau et al., 2004). Thus even if the follow-up is not performed until the target field is almost setting, we can still reject some candidates spectroscopically, even when photometric follow-up is not useful. For periods of a few days the limiting value for the mass detectable with these instruments is about 5 10M Jup. The spectra obtained with these instruments also allow us to characterize the host star. We use a library of synthetic spectra to derive values for the effective temperature and surface gravity (assuming solar metallicity) and also the line broadening. In our experience, rotational broadening of more than 10 km s 1 is a strong hint that the companion is a star, with enough tidal torque to synchronize the rotation of the host star with the orbital motion. Although the gravity determination is relatively crude, with an uncertainty of perhaps 0.5 in log g, it is still very useful for identifying those host stars that are clearly giants with log g 3.0. We presume that these stars must be the third member of a system that includes a main-sequence eclipsing binary, either a physical triple or a chance alignment, and we make no further follow-up observations. Our spectroscopic classification of the host star is a first step toward estimating the stellar mass and radius. These estimates, in turn, may be combined with the observed radial velocity variation and light curve to yield estimates of the mass and radius of the companion. Although the use of follow-up spectroscopy has the scheduling advantages outlined

26 above, the combination of both spectroscopy and photometry may be needed in the case of a blend. Such a candidate might pass our spectroscopic test as a solitary star with constant radial velocity, if the eclipsing binary of the triple is faint enough relative to the primary star (as was the case for GSC 03885-00829; O Donovan et al. 2006b, see chapter 3). High-precision, high-signal-to-noise spectroscopic observations of the few remaining candidate transiting planets should reveal the mass (and hence true nature) of the transiting companion. However, even after a spectroscopic orbit implying a planetary companion has been derived, care must be taken to show that the velocity shifts are not due to blending with the lines of an eclipsing binary in a triple system (e.g., Mandushev et al., 2005). It may be hard to see the lines of the eclipsing binary, partly because the eclipsing binary can be quite a bit fainter than the bright third star, and partly because its lines are likely to be much broader due to synchronized rotation. In some cases it may be possible to extract the velocity of one or both the stars in the eclipsing binary using a technique such as TODCOR (Mandushev et al., 2005). Combining modeling of the photometric light curve and information from the spectroscopic pseudo orbit for the system can help guide the search for the eclipsing binary lines. Even if the lines of the eclipsing binary cannot be resolved, a bisector analysis of the lines of the third star may reveal subtle shifts that indicate a binary companion. Follow-up observations with one-meter class telescopes both remove astrophysical false positives from consideration and prepare for the eventual modeling of newly discovered transiting planets. In the case of our field in Andromeda, our follow-up ruled out all of our planet candidates, and provided us with a variety of false positives to study.

27 Figure 2.1 Transit visibility plot for the Andromeda field calculated from observations made using Sleuth alone (light gray), Sleuth and the PSST (black), and all three TrES telescopes (dark gray). The fraction of transit signals with a given period identifiable from the data is plotted, assuming a requirement of observing three distinct transit events, with coverage of at least half of each individual event. About 80% of transit events with periods less than 8 days should be recoverable from the TrES observations, whereas the Sleuth observations alone provide 80% coverage only up to five-day periods.

28 2.3 Initial Observations of a Field in Andromeda with the TrES Network In 2003 August, we selected a new field centered on the guide star HD 6811 (α = 01 h 09 m 30.13 s, δ = +47 14 30.5 J2000.0). We designated this target field as And0, the first TrES field in Andromeda. We observed this field with each of the TrES telescopes. Although the TrES network usually observes concurrently, in this case weather disrupted our observations. Sleuth monitored the field through an SDSS r filter for 42 clear nights between UT 2003 August 27 and October 24. STARE began its observations with a Johnson R filter on UT 2003 September 17 and observed And0 until UT 2004 January 13 during 23 photometric nights. PSST went to this field on UT 2003 November 14 and collected Johnson R observations until 2004 January 11, obtaining 19 clear nights. We estimate our recovery rate for transit events should be 100% for orbital periods P < 6 days, declining to 70% for P = 10 days (see figure 2.1), where here our recovery criterion is the observation of at least half the transit from three distinct transit events. We note that this recovery rate is a necessary but not sufficient criterion to detect transiting planets, since it neglects the signal-to-noise ratio and the detrimental effect of non-gaussian noise on it (see, e.g., Gaudi, Seager, & Mallen-Ornelas, 2005; Gaudi, 2005; Pont, Zucker, & Queloz, 2006; Smith et al., 2006; Gaudi & Winn, 2007). We used an integration time of 90 s for our exposures. During dark time, we took multicolor photometry (SDSS g, i, and z for Sleuth and Johnson B and V for PSST and STARE) for stellar color estimates. 2.4 Searching for Transit Candidates in Andromeda We have previously described in detail our analysis of TrES data sets in Dunham et al. (2004) and O Donovan et al. (2006b, see chapter 3 for more details). We summarize

29 Table 2.1. TrES labels, 2MASS and GSC designations, and approximate V magnitudes for And0 candidate transiting systems Candidate 2MASS a GSC b V T-And0-00948 01083088+4938442 03272-00845 11.4 T-And0-01241 00531053+4717320 03266-00642 11.6 T-And0-02022 01023745+4808421 03267-01450 12.0 T-And0-02462 01180059+4927124 03272-00540 12.2 T-And0-03874 00545421+4805505 03266-00119 12.7 T-And0-03912 00595445+4902030 03271-01102 12.7 a Designations from 2MASS Catalog (Cutri et al., 2003), giving the coordinates of the sources in the form hhmmss.ss+ddmmss.s J2000.0. b GSC Catalog (Lasker et al., 1990). here the analysis for this field. We used standard IRAF 4 (Tody, 1993) tasks or customized IDL routines to calibrate the images from the three telescopes. For each telescope, we derived a standard list of stars visible in the images from this telescope and computed the corresponding equatorial coordinates using the Tycho-2 Catalog (Høg et al., 2000b). We applied differential image analysis (DIA) to each of the separate photometric data sets from the three telescopes using the following pipeline based in part on Alard (2000). For each star in our standard star lists, we obtained a time series of differential magnitudes with reference to a master image. We produced this master image by combining 20, 18, and 15 of the best-quality interpolated images in our Sleuth, PSST, and STARE data sets, respectively. Since small-aperture, widefield surveys such as TrES often suffer from systematics (caused, for example, by variable atmospheric extinction), we decorrelated the time series of our field stars. Initially, we examined our Sleuth observations separately, as these dominate the TrES data set, providing 50% of the data. We binned the Sleuth time series using 4 IRAF is distributed by the National Optical Astronomy Observatories, which are operated by the Association of Universities for Research in Astronomy Inc. under cooperative agreement with the National Science Foundation.

30 Table 2.2. Transit properties for the six TrES And0 candidates Candidate SDE Depth Period Duration N a Telescope(s) b True Nature (mag) (days) (hr) T-And0-00948 19.3 0.005 1.1198 1.6 7 S,T Eclipsing binary T-And0-01241 9.6 c 0.009 4.6619 3.4 2 S A-type star T-And0-02022 13.8 0.017 4.7399 3.4 5 S,P,T Eclipsing binary T-And0-02462 21.5 0.019 3.0691 2.2 3 S,P Eclipsing binary T-And0-03874 12.7 0.007 2.6540 3.2 7 S,P Blend T-And0-03912 18.2 0.007 2.3556 3.4 7 S,P,T Rapidly rotating A-type star a The number of distinct transits observed in the TrES data set. b TrES telescopes that detected transits of this candidate, where S is Sleuth, P is PSST, and T is STARE. c Here the SDE is based on the Sleuth data set rather than the TrES combined observations.

31 Figure 2.2 Calculated rms residual of the binned data versus approximate V magnitude for the stars in our TrES And0 data set. The number of stars with rms below 1%, 1.5%, and 2% are shown.

32 Figure 2.3 Light curves of the six TrES candidates from the And0 field in Andromeda. The labels denote the source of the light curve. The timeseries have been phased to the best-fit period identified by the box-fitting algorithm of Kovács et al. (2002) using the TrES data, with the exception of T-And0-01241, whose period was derived from the Sleuth data. The transit event is not present in the data from the other telescopes gathered at the same orbital phase.

33 nine-minute bins to reduce computation time. The rms scatter of the binned data was below 0.015 mag for approximately 7800 stars. We searched the time series for periodic transit-like dips using the box-fitting least-squares transit-search algorithm (BLS; Kovács, Zucker, & Mazeh 2002, see appendix C), which assigns a Signal Detection Efficiency (SDE) statistic to each star based on the strength of the transit detection. We restricted our search to periods ranging from 0.1 to 10 days. Having sorted the stars in order of decreasing magnitude and decreasing SDE, we visually examined each stellar light curve (phased to the best-fit period derived by the algorithm) in turn, until we determined that we could no longer distinguish a transit signal from the noise. We identified six transit candidates (see Tables 2.1 and 2.2, and figure 2.3). We then combined the three TrES data sets, which optimized our visibility function (see figure 2.1), and allowed us to confirm the detection of real eclipse events using simultaneous observations from multiple telescopes. We produced the combined TrES data set as follows. For each star in the Sleuth standard star list, we attempted to identify the corresponding stars in the other two lists. We computed the distances between a given Sleuth standard star and the PSST standard stars, and matched the Sleuth star with a PSST star if their angular separation was less than 5 (0.5 pixels). Because of the slight differences in the chosen filter and field of view, some Sleuth stars did not have corresponding PSST stars. We created a new standard star list from the Sleuth and PSST star lists, with only one entry for each pair of matched stars and an entry for each unmatched star. We then repeated this procedure with this new star list and the STARE list to produce the TrES field standard star list. For each star, we then combined the relevant time series, chronologically reordered the data, and binned the data. The rms scatter of the averaged TrES data was below 0.015 mag for 9148 stars (see figure 2.2) out of the 29,259 stars in the field. We repeated the BLS transit search, but did not identify any new candidates. From figure 2.1, we can see that a visibility of 80% had already been achieved for P < 5 days for the Sleuth data alone, and the addition of the STARE and PSST data did not significantly increase the detection space for those short periods. The lack of additional candidates with

34 these orbital periods is not surprising. However, for longer periods, the visibility for the TrES network is much better than for the single Sleuth telescope, yet we did not find new candidates with these periods. We proceeded to our follow-up observations of these six candidates with larger telescopes. 2.5 Follow-up of Candidate Transiting Planets Many of the bright stars within our field were also observed as part of other surveys. We identified our candidates in online catalogs and compared these observations with our expectations based on the planet hypothesis. We found Tycho-2 (Høg et al., 2000a,b) visible (B T V T ) colors for two of our candidates, and Two Micron All Sky Survey (2MASS; Cutri et al. 2003) infrared (J K s ) colors for all six (see Table 2.3). We searched the USNO CCD Astrograph Catalog (UCAC2; Zacharias et al. 2004) for the proper motions of the stars. All of the candidates display a measurable proper motion, consistent with nearby dwarfs. However, these proper motions were not sufficiently large to rule out distant, high-velocity giants. Finally, we retrieved Digitized Sky Survey 5 (DSS) images of the sky surrounding each candidate to check for possible nearby stars of similar brightness within our PSF radius. None were found. We observed the six And0 candidates starting on UT 2004 September 28 using the Harvard Smithsonian Center for Astrophysics (CfA) Digital Speedometers (Latham, 1992). These spectrographs cover 45Å centered on 5187Å and have a resolution of 8.5 km s 1 (a resolving power of λ/ λ 35,000). We cross-correlated our spectra against a grid of templates from our library of synthetic spectra to estimate various stellar parameters of our targets and their radial velocities. J. Morse computed this spectral library, using the Kurucz model atmospheres (J. Morse & R. L. Kurucz, 2004, private communication). Assuming a solar metallicity, we estimated the effective temperature (T eff ), surface gravity (g), and rotational velocity (v sin i) for each 5 The Digitized Sky Survey (see http://archive.stsci.edu/dss/) was produced at the Space Telescope Science Institute under US Government grant NAG W-2166. The images of these surveys are based on photographic data obtained using the Oschin Schmidt Telescope on Palomar Mountain and the UK Schmidt Telescope. The plates were processed into the present compressed digital form with the permission of these institutions.

35 Table 2.3. Photometric and spectroscopic properties of the six TrES And0 candidates Candidate vr a P(χ 2 ) b Teff c log g c v sin i c µ d BT VT e J Ks f (km s 1 ) (K) (km s 1 ) (mas yr 1 ) (mag) (mag) T-And0-00948 28.82 ± 0.53 0.180 5250 3.0 3 3.0 0.87 0.60 T-And0-01241 13.05 ± 4.06 9500 4.5 55 7.7 0.07 0.01 T-And0-02022 4.47 ± 27.87 0.000 7000 3.5 22 3.5 0.24 T-And0-02462 11.31 ± 4.73 0.008 6250 3.5 77 5.5 0.14 T-And0-03874 15.41 ± 0.31 0.747 5500 3.5 2 6.0 0.66 T-And0-03912 35.26 ± 3.43 7750 3.5 88 8.2 0.25 a The mean radial velocity. b The probability that the observed χ 2 should be less than a value χ 2, assuming that our model of a star without radial velocity variation is correct. c For a discussion of the errors in these spectroscopic data, see section 2.6. d UCAC2 proper motions (Zacharias et al., 2004). e Tycho-2 visible colors (Høg et al., 2000a,b). f 2MASS infrared colors (Cutri et al., 2003).

36 Table 2.4. Details of spectroscopic observations of the six TrES And0 candidates Candidate Time of Observation Photometric Orbital Phase Radial Velocity HJD km s 1 T-And0-00948 2453276.8749 0.46 28.16 ± 0.36 2453277.8530 0.33 29.19 ± 0.40 2453278.8236 0.20 29.00 ± 0.32 T-And0-01241 2453276.8062 0.73 13.05 ± 4.06 T-And0-02022 2453276.8642 0.80 35.46 ± 0.74 2453301.8475 0.07 14.14 ± 0.99 2453334.7734 0.02 3.77 ± 0.94 2453548.9711 0.21 39.63 ± 1.12 2453575.9740 0.90 24.54 ± 1.05 2453576.9683 0.11 20.34 ± 1.16 2453626.9029 0.65 23.98 ± 1.13 2453627.9043 0.86 29.37 ± 0.70 2453628.8392 0.06 9.31 ± 0.81 2453629.8518 0.27 42.28 ± 1.12 2453630.8361 0.48 15.29 ± 1.24 2453631.8335 0.69 30.93 ± 0.86 2453632.8135 0.90 24.80 ± 0.87 2453633.8047 0.10 20.66 ± 1.02 2453636.8830 0.75 37.44 ± 0.86 2453779.5864 0.84 29.64 ± 1.14 candidate (see Table 2.3) from the template parameters that gave the highest average peak correlation value over all the observed spectra.

37 Table 2.4 (cont d) Candidate Time of Observation Photometric Orbital Phase Radial Velocity HJD km s 1 T-And0-02462 T-And0-03874 2453276.8864 0.11 7.55 ± 1.40 2453686.7688 0.18 14.50 ± 3.74 2453276.8192 0.17 15.81 ± 0.44 2453277.8381 0.55 15.15 ± 0.41 2453301.8342 0.59 15.09 ± 0.43 2453334.7619 0.00 15.50 ± 0.43 T-And0-03912 2453276.8398 0.22 35.26 ± 3.43 For the three candidates with low stellar rotation, v sin i < 50 km s 1, we obtained several spectra over different observing seasons to determine the radial velocity variation of each star. Table 2.4 details our spectroscopic observations. For these slowly rotating candidates, the typical precision for our spectroscopic parameters is T eff = 150 K, log g = 0.5, v sin i = 2 km s 1, and V = 0.5 km s 1. The precision of the estimates degrades for stars with large v sin i values or few spectroscopic observations. We obtained high-precision photometry of T-And0-03874 on UT 2004 December 19 using the Minicam CCD imager at the FLWO 1.2-meter telescope on Mount Hopkins, Arizona. Minicam consists of two 2248 4640 pixels thinned, backsideilluminated Marconi CCDs mounted side-by-side to span a field of approximately 20.4 23.1 bisected by a narrow gap. We employed 2 2 pixels binning for an effective plate scale of 0.6 pixel 1 and read out each half CCD through a separate amplifier. We offset the telescope to place T-And0-03874 centrally on one amplifier region and autoguided on the field. We obtained concurrent light curves in three filters by cycling continuously through the SDSS g, r, and z filters with exposure times of 90, 45, and 90 s respectively. The seeing was poor (FWHM 3 7 ) and varied throughout our observations. Unfortunately, high winds forced us to close the dome during the night and we obtained only partial coverage of the predicted event. Late

38 in the scheduled observations, we re-opened for a short time. We subtracted the overscan bias level from each image and divided each by a normalized flat field constructed from the filtered mean of twilight sky exposures. To construct a light curve of T-And0-03874 and neighboring bright stars, we located the stars in each image. We measured stellar fluxes in a circular aperture and subtracted the sky as estimated by the median flux in an annulus centered on the star (iteratively rejecting deviant sky pixel values). We used a relatively large 12 radius aperture and sky annulus with inner and outer radii of 15 and 27 respectively in an effort to reduce systematic errors arising due to the poor and variable seeing conditions. We first corrected the flux of each star by an amount proportional to its air mass in each exposure by using extinction coefficients for each filter based on previous experience with Minicam photometry. Second, we selected a group of bright, uncrowded stars near T-And0-03874 as potential comparison stars. In each exposure we calculated the mean flux of the comparison stars weighted according to brightness. We assumed that any variations in this mean flux represented extinction in each image and used them to apply corrections to each light curve. We then inspected by eye the light curve of each comparison star and fit the light curves to models of constant brightness to find χ 2 statistics. We removed from our group of comparison stars any star that showed significant variations. We recalculated the extinction corrections iteratively in this manner until we achieved no variation in the comparison star light curves. We accepted 29, 32, and 4 comparison stars for the g-, r-, and z-band light curves respectively. Finally, we normalized the flux in the light curves of T-And0-03874 with respect to the out-of-eclipse data. 2.6 Rejecting False-Positive Detections Based on our detailed investigations of the candidates, we eliminated each And0 candidate as follows. In the case of T-And0-00948, the TrES light curve (figure 2.3) shows a secondary eclipse. The low surface gravity (log g = 3.0) is that of a distant giant star, consistent

39 with the red color J K s = 0.60 mag (since the majority of stars with J K s > 0.5 mag are expected to be giants, see, e.g., Brown, 2003). There was no observed variation in the radial velocity of this candidate (v r = 28.82 km s 1 ). T-And0-00948 is most likely the primary star of a diluted triple system. Upon further examination of the individual light curves for T-And0-01241, we noticed that only Sleuth had observed transit events for this system. Neither PSST nor STARE had observed this field during the time of the Sleuth transit events, preventing a comparison of the light curves. Based on the Sleuth data, we predicted the times of transits during the entire TrES And0 campaign. STARE did observe T-And0-01241 at a time at which it was predicted to transit but did not observe the transit. It was therefore possible that this was a photometric false positive. However, we did not pursue this further, as we obtained sufficient evidence from the follow-up spectroscopy to discount the system. A dwarf with log g = 4.5, the star has the high effective temperature (T eff = 9500 K) and blue colors (J K s = 0.01 mag) of an early A star. Figure 2.4b shows the nearly featureless spectrum of this star. For such a large star with a radius R 2.7R, the observed transit depth of 0.9% indicates a non-planetary size (R = 2.5R Jup ) for the eclipsing body. The radial velocity of T-And0-02022 varies with an amplitude corresponding to a stellar-mass companion. We determined this system to be an eclipsing binary, comprising a slightly evolved F dwarf and an M dwarf. This system has a mass function f(m) = 0.0304±0.0013 M and an eccentricity of 0.027±0.014. Assuming an orbital inclination of i 90 and a mass of 1.6 M consistent with the effective temperature (T eff = 7000 K), we estimated the mass of the companion to be m = 0.5 M. Figure 2.5 shows the radial velocity orbit. The circular orbit of T-And0-02022 allows us to constrain the stellar radius, independent of our derived spectral type and luminosity class. The circular orbit implies orbital synchronization and orbital-rotation axes alignment, since circularization has the longest timescale of these processes. (This should apply except for very low mass secondaries; see Hut, 1981 for a derivation of these timescales for close binary systems and a comparison for different binary mass ratios and moments of inertia of the primary star.) We can therefore assume

40 (a) (b) (c) (d) (e) (f) Figure 2.4 Sample spectra of the And0 TrES candidates obtained with the CfA Digital Speedometers on the FLWO 1.5-meter telescope: (a) T-And0-00948 has the low surface gravity of a giant star, (b) T-And0-01241 has the featureless spectrum of an A-type star, (c) T-And0-02022 is an evolved F dwarf, (d) T-And0-02462 and (e) T-And0-03912 display broadened lines due to the rapid rotation of these stars, and (f) T-And0-03874 is an early K-type dwarf.

41 40 20 0-20 -40 0.2.4.6.8 1 ORBITAL PHASE Figure 2.5 Radial velocity orbit of T-And0-02022 as determined from our Digital Speedometer spectra. This system was quickly rejected as a candidate transiting planet after the large radial velocity variation was determined from initial spectroscopic observations. Additional epochs produced a precise orbit with an eccentricity of e 0.03. An initial mass estimate of 0.5 M was derived for the companion. Future photometry will lead to more precise mass determination for both component stars (see text for a discussion).

42 that the stellar rotation period is the same as the orbital period of 4.7399 days. We then use the observed rotational broadening (v sin i = 22 km s 1 ) to estimate the radius of the star to be 2.0 R. Future photometric observations of this systems with KeplerCam (Szentgyorgyi et al., 2005) are planned to more precisely measure the eclipse depth and to derive the radius and true mass of each star in this binary (J. M. Fernández et al. 2007, in preparation). T-And0-02462 is a rapid rotator with v sin i = 77 km s 1 and displays rotationally broadened lines (see figure 2.4d), limiting the possibility of detecting the radial velocity variation caused by a planet. Regardless, we inferred a binary nature for this candidate using the combined TrES data, which proved essential in identifying this false positive. Our initial Sleuth light curve of T-And0-02462 showed no evidence of a stellar-mass companion, and we derived a BLS best-fit period of 1.5347 days. On examining the TrES light curve of T-And0-02462 (see figure 2.3), we noticed a secondary eclipse. Also, the BLS best-fit period for our TrES observations of T-And0-02462 is twice that derived from the Sleuth data alone. When we reexamined the Sleuth data phased to the TrES period, the secondary eclipse is not visible due to inadequate coverage at that phase. This resulted in a derived period for the Sleuth data half that of the true period. The red color of T-And0-03874 (J K s = 0.66 mag) and the effective temperature (T eff = 5500 K) calculated from the spectrum shown in figure 2.4e are consistent with an early K-type star. The radial velocity of T-And0-03874 was observed to remain constant at 15.53 km s 1. However, the low estimated surface gravity (log g = 3.5) suggested this star is a giant star and part of a diluted triple. The photometric follow-up (see figure 2.6) failed to recover transits of T-And0-03874, but did observe a nearby eclipsing binary T-And0-02943 undergoing a deep eclipse at the predicted transit time (also shown in figure 2.6). When we examined our TrES observations for T-And0-02943, we saw that the period of this eclipsing binary was that originally derived for T-And0-03874, namely 2.654 days. This eclipsing system lies 45 away from T-And0-03874, comparable to the PSF radius of our TrES aperture photometry (30 ; 3 pixels). The angular resolution ( 1 pixel 1 ) of the 1.2-meter photometry is

43 Figure 2.6 Follow-up g-, r-, and z-band photometry with the FLWO 1.2-meter telescope of T-And0-03874 (squares) and a neighboring star (diamonds), designated T-And0-02943, that lies 45 away. (The inset 2 2 Digitized Sky Survey image shows T-And0-03874 at the center and T-And0-02943 above, to the north.) The flux from each star has been normalized using the out-of-eclipse data, and an offset applied for the purpose of plotting. Inclement weather prevented complete coverage of the predicted eclipse event. Nevertheless, it is evident that T-And0-02943 displays a deep (>14%) eclipse, whereas the flux from T-And0-03874 is constant. The observed apparent transits of T-And0-03874 were the result of the blending of the light from these two systems. With a FWHM for T-And0-02943 of 2.5 pixels ( 25 ), some of the light from this star was within the photometric aperture radius (3 pixels; 30 ) of T-And0-03874.